KR102154667B1 - Pattern measuring device, and computer program - Google Patents

Abstract

An object of the present invention is to provide a pattern measurement device that enables selection of device conditions for obtaining an appropriate deviation or estimation of an appropriate deviation. In the present invention, based on a plurality of measured values acquired by a charged particle beam device, a pattern measuring device provided with an arithmetic processing device for obtaining a deviation of a measured value of a pattern to be measured, a plurality of measured values formed at different positions A pattern measurement device characterized by obtaining σ _SEM0 representing the measurement reproducibility error using the deviation of (σ _measured ) and σ ² _observed ＝σ ² p _attern0 ／Np+σ ² _sem0 ／(Np·Nframe) is proposed. do. σ _pattern0 is the deviation due to the shape error of the pattern, Np is the number of measurement points, and Nframe is a value that changes according to the device conditions.

Description

Pattern measuring device, and computer program

The present disclosure relates to an apparatus for measuring patterns formed on a semiconductor device, etc., and a computer program for executing measurement processing on a computer, and in particular, measures index values of patterns such as variations in pattern shape with high accuracy, regardless of variations in measurement errors, etc. It relates to a pattern measuring device and a computer program.

With the miniaturization of semiconductor devices, the importance of measurement and inspection using a Scanning Electron Microscope (SEM) capable of visualizing a pattern having a width on the order of nanometers is increasing. In Patent Document 1, an evaluation method for evaluating the unevenness formed on the pattern sidewall using a signal obtained by SEM is described. These irregularities are referred to as line edge roughness (LER) or line width roughness (LWR). Since roughness affects the characteristics of the circuit formed on the semiconductor device, it is necessary to perform an appropriate evaluation. In addition, Patent Document 2 describes an evaluation method for evaluating the shape of a pattern by comparing a reference pattern generated using design data with an outline extracted from an SEM image. In Patent Document 2, an evaluation method (Edge Placement Error: EPE) for evaluating a deviation between the design data and the corresponding point of the outline data by measuring the distance between the corresponding points is described.

Japanese Patent No. 4881427 (corresponding US Patent USP 7,619,751) Japanese Patent No. 4887062 (corresponding US patent USP8,019,161)

As described above, with the miniaturization and complexity of semiconductor patterns, the importance of two-dimensional shape measurement such as roughness evaluation and shape evaluation using a plurality of EPE measurement results is increasing. On the other hand, the measured value obtained by SEM varies due to various factors. For example, even if a sampled pattern is accurately measured for measurement, it is only an accurate measurement of the pattern to be sampled, and even if process control of a manufacturing device is performed based on this measurement result, appropriate control is performed. There are cases when it cannot be done. In addition, in the case of a scanning electron microscope that detects electrons obtained based on scanning of an electron beam, for example, the emission direction of electrons emitted from the sample is random and becomes non-uniform for each measurement, so even if the same pattern is measured, It is not possible to form exactly the same signal waveform, and as a result, there is a possibility that measurement results for performing appropriate process control may not be obtained.

As described above, there are measurement deviations inherent in devices such as SEM, and the index values representing the pattern shape include deviations due to variations in measurement results (variations in measurement errors) and true shape changes (shape deviations) of the pattern to be evaluated. ) Are mixed, and it may become difficult to evaluate the shape of an appropriate pattern. In Patent Documents 1 and 2, there is no mention of reducing the influence on the measurement result of the variation caused by such variation in measurement error.

Hereinafter, a pattern measuring apparatus and a computer program for at least one of selection of an apparatus condition for obtaining an appropriate deviation and estimation of an appropriate deviation are proposed.

As an aspect for achieving the above object, a pattern measurement device including an arithmetic processing device that calculates a deviation of a measured value of a pattern to be measured based on a plurality of measured values acquired by a charged particle beam device below, comprising: _{Characteristic} to obtain σ _SEM0 representing the measurement reproducibility error using the deviation (σ _observed ) of a plurality of measurement values formed at different locations and σ ² _observed ＝σ ² _pattern0 ／Np+σ ² _sem0 ／(Np·Nframe) We propose a pattern measuring device to be used. σ ² _pattern0 is the deviation due to the shape error of the pattern, Np is the number of measurement points, and Nframe is a value that changes according to the device conditions.

According to the above configuration, it becomes possible to select an apparatus condition for obtaining an appropriate deviation or to estimate an appropriate deviation.

1 is a diagram showing a relationship between a deviation specified by pattern measurement and a cause of the deviation.
Fig. 2 is a diagram showing a relationship between signal electrons emitted by incidence of a beam on a pattern and a pattern shape.
Fig. 3 is a diagram showing a relationship between a pattern length (for example, a length in an edge direction of a line pattern) and a signal deviation.
Fig. 4 is a diagram showing a relationship between a deviation specified by pattern measurement and a cause of deviation.
5 is a diagram showing an example of a semiconductor measurement system.
Fig. 6 is a diagram showing an outline of a scanning electron microscope.
7 is a diagram showing an example of a semiconductor wafer to which a resist pattern is transferred.
Fig. 8 is a diagram showing a state in which measurement points are assigned to each of a plurality of chips of a semiconductor wafer.
Fig. 9 is a diagram showing a measurement area (measurement area) allocated within a field of view of a scanning electron microscope.
Fig. 10 is a flow chart showing a measurement processing step by SEM.

In the manufacturing process of electronic components such as semiconductor integrated circuits and high-frequency devices, a substrate such as a semiconductor or various thin films on the substrate is used to connect electronic components such as a plurality of transistors and memory elements or the elements using lithography and etching techniques. Create patterns such as circuits. In state-of-the-art devices that are undergoing miniaturization, the dimensions of each element and circuit reach 10 nm, and are expected to be further reduced in the future. Since the performance of the electronic component largely depends on the planar shape of the device or circuit, it is preferable that the planar shape is generally formed with an accuracy of about 1/10 to 1/3 of the design dimension of the pattern.

However, due to various factors, a deviation occurs between the design pattern and the pattern of an element or circuit actually manufactured in reality. Such a shift can be handled quantitatively by evaluating the shift (EPE) of the edge of the pattern of an element or circuit that is actually manufactured and the location of the edge where the pattern should originally exist. The factors that cause EPE are classified into the following three.

These three are: (1) dimensional change such as deviation of the pattern average dimension, overall thickness or thinness, (2) deviation of the pattern average or center position, overall position deviation, and (3) the pattern edge or local pattern size. It is a natural change.

(1) is caused by thinning or thickening of the size of the resist pattern transferred by the exposure apparatus, or by side etching, and can be corrected by adjusting exposure conditions such as exposure amount and focus position, or process conditions such as etching. There is. (2) is caused, for example, by a misalignment between the underlying pattern and the mask transfer image by the exposure apparatus, and can be corrected by alignment adjustment of the exposure apparatus. (3) is a random or local edge position variation that is difficult to control by the above, for example, in the intermediate region of the exposure intensity near the edge of the projected image, caused by a stochastic variation in a chemical reaction or dissolution phenomenon in the material. I think it is. The amplitude of the deviation, for example, reflects the width of the intermediate exposure range = the steepness of the exposure profile (defocus condition, etc.), so it varies depending on the effective projection image contrast, and can be controlled to some extent. .

The deviation of the above (1) to (3) varies due to spatial and temporal fluctuations, and the influence on the device performance depends on the relationship between the spatial period of the fluctuation and the device dimensions. When the fluctuation period is smaller than the device dimensions, the fluctuation degrades the device performance itself. On the other hand, when the variation period is larger than the element dimensions, the variation becomes element performance variation. In general, the variable spatial period of (3) is smaller than that of (1) and (2), but with miniaturization, the effect of (3) with a small spatial period on the device performance variation becomes present. have. Fig. 3A shows the deviation of the average width of line patterns having different lengths. The shorter the line length of interest, the greater the variation in the line width due to (3).

In order to suppress the shift caused by the above manufacturing process and material, it is required to correctly measure the pattern size and shape, accurately grasp the distribution in the wafer surface and local dimensional variation, and correct the process condition and the device knob.

In order to suppress discrepancy caused by the manufacturing process and material as described above, the pattern dimensions and shape are correctly measured by SEM, the distribution within the wafer surface and local dimensional variations are accurately grasped, and the process conditions and the device knob are corrected. It is required to do. For this, measurement reproducibility in pattern measurement using SEM is important.

However, there is a reproducibility error in pattern measurement by SEM. The intrinsic factor of reproducibility error is the particle nature of electrons. In Fig. 2, the principle of edge position measurement of a line pattern by SEM is schematically shown. In general, a focused electron beam is scanned in a direction perpendicular to the edge direction (y direction) (x direction) to detect secondary electrons or backscattered (reflected) electrons emitted from the sample, and the irradiation position of the electron beam (x coordinate ) And the detection signal to obtain a signal strength profile.

Since the signal strength becomes maximum when incident near the edge, the edge position can be defined, for example, by setting an appropriate threshold for the relative signal strength. However, when detecting secondary electrons or reflected electrons, each process, such as the incident position of the electrons, the direction of scattering and reflection inside the sample, and generation of the secondary electrons, is a probability process, and as a result, the signal intensity profile becomes uneven. Therefore, the edge position obtained from the profile is also non-uniform. According to the central limit theorem, if the statistics of a plurality of sample groups formed by extracting n measurement objects from the parent group of the mean value (A) and variance (σ ² ) are calculated, the average value of the plurality of sample groups The average of is A, and the variance is σ ² /n. Therefore, the measurement of the deviation (σ_) of the result (average value of the distribution) obtained when repeatedly measuring exactly the same object by a measuring device having a normal distribution of the standard deviation (σ) as a reproducibility error is σ_measurement = σ/ Becomes √n. Therefore, measurement reproduction accuracy is improved by increasing the number of measurements n.

However, for several reasons as described below, in practice, objects of the same design shape formed in different places are often measured. First of all, if the measurement is repeated at the same location, there is a possibility that the measurement result may change due to sample damage (contamination or shrinkage). In the case where both the suppression of contamination or shrinkage and the increase in the number of measurements for high-precision measurement are compatible, it is necessary to change the position. Secondly, even with the same design pattern in general, the actual shape becomes uneven depending on the location. In this case, even if a complete measurement is performed for any one pattern, it is only that the true value at one point of the deviation has been obtained. Therefore, in order to obtain the statistical average of the design shape of the measurement object, it is necessary to measure a plurality of samples and obtain the average. Third, it is often desired to clarify the magnitude of the deviation of the shape of the same design pattern according to the above location, or the distribution of shape change itself. Variations depending on the location vary from inter-wafer variation, intra-wafer distribution, and intra-chip distribution to dimensional variation in local regions. In this case, it goes without saying that measurement is performed at a plurality of locations.

In the semiconductor manufacturing process, the process device is controlled based on the change in the average shape of the target pattern, but in this way, when measurements are performed multiple times at different locations, the deviation of the measurement result is the difference due to the measurement reproducibility error and the true There are cases where shape deviations are mixed, and it is not known whether the change in the measurement result reflects the change in the true pattern shape.

In addition, when measuring dimensions, shapes, and local deviations thereof at multiple locations within the wafer surface or within the chip, and controlling the process equipment so that their distribution falls within the allowable range, the measured abnormal shape and actual In some cases, it is not known whether the difference in the pattern reflects the difference in the true shape. In either case, in order to perform reliable process control, it is required that the reproducibility error at the time of repeated measurement is sufficiently small compared to the difference in measurement results caused by a true variation in the shape of the measurement object.

In the examples described below, in order to quantitatively handle variations and the like, a plurality of feature quantities are modeled as follows. In addition, in this embodiment, a number of identical design patterns existing in a predetermined area on the wafer are considered as a population. The n patterns are sampled from the population, their dimensions and shapes are measured, and an average value and a deviation are obtained from the distribution of the measurement results as a characteristic quantity. The characteristic quantities handled in the examples described below are the average of the average value of the measurement results (CD center value), the variance of the average value of the measurement results (measurement reproducibility error of the CD center value), and the average value of the variance of the measurement results (CD It is four of the average of the deviation = CDU or LCDU), and the variance of the variance of the measurement result (measurement reproducibility error of CDU or LCDU). In addition, CD is a measurement value of the line width of a pattern, for example. In order to obtain a true shape deviation, it is required to bring the CD center value and CDU·LCDU as close as possible to that of the population, or to bring the measurement reproducibility error as close as possible to zero.

The measurement result output from the measurement device becomes uneven due to two factors: a deviation (σ_pattern) caused by a variation in the shape of the target pattern itself, and a deviation (σ_sem) caused by a measurement reproducibility error. Here, since the two are independent of each other, the variance (σ ² _observe) observed as a result of the measurement is the sum of the variances due to both. That is, it can be expressed as σ ² _observed = σ ² _pattern+σ ² _sem.

Hereinafter, a method of accurately estimating the σ_pattern of the population from σ_observed by separating the cause of the deviation into two terms will be described.

In the examples described below, an index value representing a true deviation is calculated mainly through the following steps. First, the number of measurement points (or the number of independent detection edge points in the beam scanning/measurement area) and the number of measurements per measurement point (or beam scanning/measurement conditions) are variously changed, and multiple times are measured for each device condition such as the number of beam scanning. By repeating, the deviation distribution of the obtained measurement result is calculated|required (process 1). Next, the dependence of the beam scanning/measurement condition (or the number of measurements) of the magnitude of the deviation, and the dependence of the number of independent detection edge points (or the number of measurement points) are obtained, and the dependence is calculated as deviation = (number of measurement points (or independent detection edge points). The term depends on the number) x (integer + the term depending on the number of measurements per measurement point (or beam scanning/measurement condition)) (Step 2). Finally, based on the above relational expression, (1) the minimum beam scanning/measurement condition required to suppress the deviation from the minimum number of detected edge points to the required range is set, or (2) the magnitude of the true deviation is determined by the ( It is specified by a term*integer depending on the number of independent edge points (Step 3).

Here, the beam scanning/measurement condition is a beam condition including the number of frames or the amount of current, and the number of independent detection edge points is given by, for example, the number of measurement edges/the length of the measurement edge/scan line interval.

Specifically, the method is, for example, in calculating the true pattern edge deviation from the observation deviation, based on a signal obtained by beam scanning, σ ² _observed, which represents the deviation of the characteristic amount of the pattern included in the sample. Σ ² _observed is obtained, and σ ² _CD is calculated by substituting σ ² _observed = σ ² _CD /#cutline+σ ² _sem /#cutline#frame. σ _CD is a deviation excluding a deviation caused by a variation in measurement error, and by outputting a parameter varying according to σ _CD obtained from σ _CD or σ _CD , it becomes possible to specify a true shape deviation. In addition, σ ² _CD is an index value representing a true deviation of the measurement object, and σ ² _sem is a deviation that changes according to the conditions of the electron microscope. In addition, #cut line is a value that changes according to the edge length, and #frame is a value that changes according to beam conditions (device conditions) such as the number of frames.

The variance of the deviation centered on the average value of the target pattern shape is inversely proportional to the size of the extracted sample, while the variance of the deviation caused by the measurement reproducibility error is not only inversely proportional to the size of the extracted sample, but also inversely proportional to the number of measurements. In the embodiment described below, the variation of the shape origin is selectively extracted using this phenomenon. More specifically, when measuring the pattern size or shape using SEM, a plurality of measurement objects having the same design existing in a predetermined local area are used as a population, and a plurality of objects (samples) are selected from among them, and a plurality of The number of times is measured, and the average and variance of the measurement results are obtained. The purpose of the examples described below is to make the average and variance of samples as close as possible to the average and variance of the population with as few samples as possible. As a result of the examination based on the experiment result of the inventor, the following relationship became clear. First, the experimental results are described.

First, a pattern was exposed on a semiconductor wafer coated with a predetermined resist through a mask including a plurality of identical patterns. At that time, exposure was performed so that the exposure regions of one shot were arranged in a matrix shape. And the pattern formed by exposure was imaged by SEM. The field of view (FOV) of the SEM is set to include a line-and-space pattern, and electron beams in a direction perpendicular to the edge (x direction) at a plurality of different positions along the pattern edge direction (y direction). By scanning, two-dimensional signal intensity distributions at a plurality of different positions were obtained.

A plurality of SEM images were obtained by changing the number of scans for the same imaging area, that is, the number of frames. Next, for each of the SEM images, a plurality of measurement boxes were set for the pattern, and the dimensions of the pattern in the measurement box were measured. At this time, conditions such as the length of the measurement box and the number of measurement boxes in the FOV were variously changed. Specifically, in each measurement box, the pattern dimensions were obtained from the signal intensity distribution along the cut line in the direction perpendicular to the edge (in the x direction) at a plurality of positions spaced a certain distance in the pattern edge direction (y direction) ( Here, you may perform an appropriate averaging process). The dimensions for all cut lines in the measurement box are averaged and output as a pattern line width in the measurement box. The statistical distribution of the line width obtained for a plurality of measurement boxes was obtained, and the standard deviation of the deviation was calculated.

An SEM image was obtained by varying the number of frames (Nframes) in a line-and-space resist pattern formed on a semiconductor wafer at a plurality of positions. Next, extracting a plurality of Np different positions along the pattern edge direction (y direction), setting a cut line in a direction perpendicular to the edge (x direction), and from the electron beam signal intensity profile obtained along each cut line The pattern dimensions were obtained, and the average value and its standard deviation were calculated from the distribution of dimensions for Np cut lines.

The average value and standard deviation represent the average dimension and LER of the pattern. This procedure was repeated to investigate the dependence of the obtained average dimension and the distribution of LER on the number of cut lines (Np) and the number of frames (Nframe). The results are shown in Fig. 1(a).

As shown in Fig. 1(a), the deviation is not dependent on the number of frames, but between the area (area 1) that depends only on the number of cut lines and the area (area 2) that depends on both the number of cut lines and the number of frames. It can be seen that it is divided into two areas. That is, the variance of the observed deviation (σ ² _observed ) is expressed by the following equation.

[Equation 1]

σ ² _observed ＝σ ² _{pattern 0} ／Np+σ ² _sem0 ／(Np·Nframe)

The region 1 corresponds to a case where the N frame is large and the second term is negligible compared to the first term, the region 2 is opposite, and the second term is dominant. In Fig. 1B, the dependence of the number of cut lines and the number of frames of the two terms of the equation is schematically shown.

Next, the meaning of the equation will be described. The reason that can be expressed by the above formula is considered as follows.

The variance of the measurement result includes both a variation in the shape of a target pattern, that is, a spatial variation, and a variation due to a measurement reproducibility error, that is, a temporal variation. The causes of variation over time include (1) stochastic variation due to particle nature of electrons (number of incident electrons, incident location, scattering process, etc.), and (2) change in measurement target due to electron-material interaction (sample damage Etc.), and (3) there is a system noise of the measurement device.

Among these, (1) is due to the principle of generation of secondary electrons as described above, and σ _Sem , which represents the degree of its contribution, is based on the number of incident electrons (Ne) per unit length in the electron beam scanning, as follows: Is displayed as

[Equation 2]

σ ² _sem ∝(1／Ne)

In addition, since the number of incident electrons (Ne) is proportional to the number of frames (Nframe) and the probe current (Ip),

[Equation 3]

σ ² _sem ∝(1／Nframe)∝(1／Ip)

Can be marked as

On the other hand, in the case of measuring objects existing in different places, the measurement results become non-uniform due to spatial variation, and at the same time, measurement is inevitably performed at different times, thus including the variation due to time. At this time, the spatial deviation of the measurement target (σ _pattern ) and the measurement deviation by time (σ _Sem ) are independent.

[Equation 4]

σ ² _observed ＝σ ² _pattern +σ ² _sem

Both depend on the number of measurement points (Np) according to the central limit theorem. In other words

[Equation 5]

σ ² = σ ² _{pattern pattern0 /} Np

[Equation 6]

σ ² _sem ＝σ ² _sem0 ／(Np·Nframe)

Becomes. therefore,

[Equation 7]

σ ² _observed ＝σ ² _{pattern 0} ／Np+σ ² _sem0 ／(Np·Nframe)

Becomes.

In addition, when the beam current Ip changes, or when the change due to system noise or the like cannot be ignored, a parameter representing their contribution is added to the above equation,

[Equation 8]

σ ² _{observed ＝} σ ² _{pattern 0} ／Np+σ ² _sem0 ／(Np·Nframe)(I ₀ ／Ip)+σ ² _noise

You may do it. However, σ _Sem0 is a coefficient with respect to the beam current I ₀ .

Further, if the number of scans is repeated, there is a possibility that the dimensional shape of the measurement object is changed due to sample damage caused by the electron beam, and the measured value may shift. Since the dimensional shift due to such damage is definite as a function of the number of scans and is not a measurement deviation, it is preferable to subtract it from the measurement result in advance.

In pattern size and shape measurement using SEM, by quantifying the degree of contribution of measurement reproducibility errors, measurement conditions that can suppress measurement errors (optimization of device conditions) and extraction of true shape evaluation values by separating measurement errors This becomes possible. For this reason, it becomes possible to control and manage various manufacturing apparatuses and manufacturing processes with high reproducibility within a realistic time range by high-precision and high reproducibility measurement using an electron beam beam, and the manufacturing yield of various semiconductor devices and electronic parts can be improved. By improving, it is possible to maintain and improve their performance.

Hereinafter, a measurement apparatus for measuring a pattern formed on a semiconductor device, a computer program for executing pattern measurement on a computer, and a storage medium storing the program will be described with reference to the drawings.

5 is a diagram showing an example of a measurement system including a pattern measurement device 502 that measures a pattern based on the measurement result obtained by the SEM 501. In addition, in the example of FIG. 5, the SEM 501 which is an imaging system, and an arithmetic processing device 505 (pattern measurement device 502) that performs measurement processing based on a detection signal are connected via a network. Although an example is shown, it is not limited to this, For example, the image processing part 618 included in the scanning electron microscope illustrated in FIG. 6 mentioned later may perform arithmetic processing as described later. In the system illustrated in Fig. 5, an SEM 501, a pattern measurement device 502 that performs pattern measurement based on the obtained signal, and design data of a semiconductor device, or layout data generated based on the design data are included. A design data storage medium 503 to be stored and an input device 504 provided with a display portion are included.

6 is a diagram showing an outline of the SEM 501. The electron beam 603 is extracted from the electron source 601 by the extraction electrode 602 and accelerated by an acceleration electrode (not shown). The accelerated electron beam 603 is narrowed by a condenser lens 604 which is a form of a focusing lens, and then deflected by a scanning deflector 605. Accordingly, the electron beam 603 scans the sample 609 in one-dimensional or two-dimensional manner. The electron beam 603 incident on the sample 609 is decelerated by the negative voltage applied to the electrode built in the sample table 608 and is focused by the lens action of the objective lens 606 to cause the sample 609 to Irradiated on the surface.

Electrons 610 (secondary electrons, backscattered electrons, etc.) are emitted from the irradiated site on the sample 609. The emitted electrons 610 are accelerated in the direction of the electron source 601 by an acceleration action based on the negative voltage applied to the electrode embedded in the sample table 608. The accelerated electrons 610 collide with the conversion electrode 612 to generate secondary electrons 611. The secondary electrons 611 emitted from the conversion electrode 612 are captured by the detector 613, and the output (I) of the detector 613 changes according to the amount of the captured secondary electrons. The luminance of the display device changes according to the change in the output I. For example, in the case of forming a two-dimensional image, the deflection signal to the scanning deflector 605 and the output I of the detector 613 are synchronized to form an image of a scanning area.

In addition, the SEM illustrated in FIG. 6 shows an example in which the electrons 610 emitted from the sample 609 are converted into secondary electrons 611 in the conversion electrode 612 and detected. The configuration is not limited, for example, a configuration in which an electron correction tube or a detection surface of the detector is disposed on the trajectory of the accelerated electron may be adopted. The control device 614 supplies necessary control signals to each optical element of the SEM in accordance with an operation program for controlling the SEM called an imaging recipe.

Next, the signal detected by the detector 613 is converted into a digital signal by the A/D converter 617 and sent to the image processing unit 618. The image processing unit 618 generates an integrated image by integrating signals obtained by a plurality of scans in units of frames. For example, in the case of integrating 8-frame images, the integrated image is generated by performing an addition average process on a pixel-by-pixel basis on signals obtained by eight two-dimensional scanning.

In addition, the image processing unit 618 includes an image memory 620 that is an image storage medium for temporarily storing digital images, and a feature amount (dimension value of the width of a line or hole, a roughness index) from the image stored in the image memory 620. A CPU 619 that calculates a value, an index value indicating a pattern shape, an area value of a pattern, a pixel position serving as an edge position, etc.) is provided.

Further, it has a storage medium 621 that stores the measured values of each pattern, the luminance value of each pixel, and the like. Overall control is performed by the workstation 622, so that necessary device operation, confirmation of detection results, and the like can be realized through a graphical user interface (hereinafter referred to as GUI). In addition, in synchronization with the scanning signal supplied to the scanning deflector 605, the image memory transfers the detector output signal (a signal proportional to the amount of electrons emitted from the sample) to the address (x, y) on the corresponding memory. It is structured to remember. The image processing unit 618 also functions as an arithmetic processing device that generates a line profile from the luminance value stored in the memory, specifies the edge position using a threshold method or the like, and measures the dimensions between edges.

The SEM performing dimensional measurement based on such line profile acquisition is referred to as CD-SEM, and is used to measure various feature quantities other than measuring the line width of a semiconductor circuit. For example, irregularities called line edge roughness exist at the edges of the circuit pattern, which causes changes in circuit performance. CD-SEM can be used to measure the LER.

LER is calculated by making the degree of deviation of the actual edge position with respect to the average position of the edge position in the x direction (eg, the scanning line direction of the electron beam) as an index value.

Hereinafter, a method of adjusting and measuring the apparatus based on the detection signal obtained by the SEM will be specifically described.

[Example 1]

In this embodiment, an example in which the number of samples and imaging conditions required to obtain a CD measurement result with required measurement reproducibility are optimized according to the present invention, and measurement reproducibility errors are separated for patterns of arbitrary length, and the true pattern size and shape An example of estimating the deviation will be described.

First, a pattern is transferred to a semiconductor wafer coated with a predetermined resist using a mask including a large number of patterns having the same shape on the design data by an exposure apparatus. At this time, as illustrated in FIG. 7, a resist pattern is formed by performing exposure so that a plurality of chips 702 including the same pattern are arranged in a matrix on the wafer 701. Next, using the SEM 501 illustrated in FIGS. 5 and 6, for N ₁ regions of the wafer on which the resist pattern was formed, the number of scans for the same imaging region, that is, the number of frames, was varied under multiple conditions. Acquire an SEM image. A recipe, which is an operation program for controlling the SEM, is stored in advance in the memory 506 or the storage medium 621, and the SEM 501 executes measurement processing according to the operation program stored in the recipe. 10 is a flow chart showing a measurement process step of SEM. The sample 609 (wafer) is introduced into the vacuum sample chamber of the SEM 501 (step 1001), and the sample table 608 and the deflector for visual field movement are controlled as necessary so that the field of view is positioned at a predetermined measurement position ( Step 1002).

FIG. 8 shows an example in which a plurality of visual fields 802 are set in a region 801 in which a pattern of the same shape is formed included in the chip 702 on the wafer 701. In the example of FIG. 8, in the design data, in the area 801 in which line patterns having the same line width are arranged (30 μm in this example), a field of view 802 at a 1 μm angle is arranged at a 2 μm pitch. Is shown. In the case of this example, since the field of view 802 having 10 vertically and 10 horizontally is arranged, N ₁ is 100. Based on the detection signal obtained in each field of view, the characteristic quantity (dimension value) of the pattern is measured. In addition, since the number of scans is changed in units of chips as described above, when the field of view is positioned at different positions, the number of frames is also set (changed) (step 1003).

Next, measurement is performed by performing beam scanning at the set number of frames at the set field of view (step 1004). In the present embodiment, the imaging region (field of view, scanning region of the electron beam) is set to include a line and space pattern of the same design as illustrated in FIG. 8. 8 shows an example in which the field of view 802 is set to a line and space pattern in which line patterns 804 and spaces 805 are alternately arranged. Since the cross-sectional view of the line pattern 804 is as shown in 807, and the edge 806 is formed to be inclined with respect to the irradiation direction of the electron beam, the amount of secondary electrons emitted is relatively larger than that of other portions due to the edge effect, and high brightness It is represented by

In the present embodiment, N ₂ measurement areas having a predetermined measurement edge length were designated in the field of view 802, and the pattern width in the measurement area was measured as follows. In addition, setting of the measurement area (measurement area) is realized by setting a plurality of measurement boxes 803 in the field of view 802. 8 shows an example in which nine measurement boxes 803 (N ₂ =9) are set in the field of view 802. In the image processing unit 507 or the image analysis unit 508, a plurality of signal intensity profiles (luminance distribution waveforms) are generated in the measurement box 803 along the x direction, and a threshold method, a linear approximation method, etc. To measure the size between peaks. In addition, in the present embodiment, for the measuring region, a cut line is set in the edge vertical direction at a plurality of N ₃ positions for a pattern at a constant interval in the edge direction (y direction), and a signal strength profile along the cut line is obtained. . At this time, if necessary, image processing such as filtering incidental to the measurement SEM may be performed. 9 is a diagram showing an example in which signal intensity profiles 905 to 908 are generated for each divided regions 901 to 904 divided by cut lines. In this embodiment, a pattern dimension is calculated by applying a predetermined algorithm to the signal intensity profile of the divided area divided by each cut line, and all (N ₃ pieces (in the example of FIG. 9) included in the measurement area are 4)) By averaging the measurement results, dimensions for each measurement area are obtained. By the above processing, one measurement value was obtained for a combination of the three parameters of the imaging area in the wafer, the measurement area in the imaging area, and the number of frames.

As described above, in order to obtain measurement results of a plurality of measurement areas for each of a plurality of different frames, movement to a plurality of field of views (FOV) or a plurality of measurement areas within the same chip (step 1005) ) And measurement (step 1004) are repeated. In addition, since measurement in a plurality of fields of view is performed for each different number of frames, stage movement to different chips or field of view movement (step 1006) can be performed using a plurality of different frames (for example, N ₁ ). Repeat until the measurement in is completed.

Next, for each frame number (Nframe) condition, the image analysis unit 508 randomly samples different number (Np) areas from the N1·N2 measurement areas obtained as described above, and the measurement values Find the mean value (CD) and standard deviation (LCDU) of the distribution of. Further, this process is repeated a plurality of times, the distribution of the mean value (CD) and the standard deviation (LCDU) of the distribution of the measurement values is examined, and the variance of the deviations (?CD, ?LCDU) is obtained.

Fig. 3B shows the dependence of the number of measurement areas (N1·N2) and the number of cut lines per area (N3) of ?CD and ?LCDU in an image acquired with the same number of frames. In addition, each axis in the figure is plotted as a natural logarithmic axis.

According to the inventor's examination, it was found that the variance of the deviation is inversely proportional only to the total number of measurement points (Np = N ₁ ·N ₂ ·N ₃ ), regardless of the conditions such as the measurement line line length. 4 shows the dependence of the number of measurement areas (N1·N2) and the number of frames (Nframe) of ?CD and ?LCDU. The numbers given to the curved portions in the drawing indicate σ _CDobserved to be described later.

As described with reference to Fig. 1, there are two areas: an area that does not depend on the number of frames but depends only on the total number of measurement points (area 1) and an area that depends on both the total number of measurement points and the number of frames (area 2). (The example in Fig. 1 corresponds to the case of N ₃ =1).

Next, the image analysis unit 508 calculates a value representing the dependence of each of σCD and σLCDU on Np and Nframe to [Equation 9] and "Equation 10".

[Equation 9]

σ ² _CDobserved ＝σ ² _{CDpattern 0} ／(N ₁ ·N ₂ )+σ ² _{CDsem0 ／} ((N ₁ ·N ₂ ·N ₃ ·Nframe)·(I ₀ ／Ip)

[Equation 10]

σ ² _LCDUobserved ＝σ ² _LCDUpattern0 ／(N ₁ ·N ₂ )+σ ² _LCDUsem0 ／((N ₁ ·N ₂ ·N ₃ ·Nframe)·(I ₀ ／Ip)

Here, σ ² _CDpattern0 is the true dimensional deviation per 1 measurement area, σ ² _LCDUpattern0 is the LCDU deviation per 1 measurement area, and σ ² _CDsem0 is 1 cut line (the total number of measurement points) and the dimensional measurement error per frame. Dispersion, σ ² _LCDUsem0 is a parameter that means the variance of the LCDU measurement error per one cut line and one frame.

σ ² _CDobserved is a value obtained by measurement, and since N ₁ , N ₂ , N ₃ , and Nframe are setting conditions, the image analysis unit 508 calculates σ ² _CDsem0 by _multiple regression analysis or the like. In the case of the present example, a result of σ ² _CDsem0 = 5 nm was obtained as a value indicating the variance of the dimensional measurement error per cut line. The image analysis unit 508 _obtains σ ² _{CDsem0 based} on the stored calculation equation (step 1007), and stores it in the memory 506 or the like for optimizing the device conditions to be described later and calculating true shape deviation (step 1008). In addition, by substituting sigma ² _CDsem0 obtained by _multiple regression analysis or the like or the above-described into Equation 9, sigma ² _CDpattern0 is also calculated and stored, and is used for optimizing device conditions described later. By _obtaining σ ² _CDsem0 for each of a plurality of device conditions (the number of frames), σ ² _{CDpattern 0} in each device condition can be obtained.

Next, a method of optimizing device conditions (measurement conditions and measurement sampling conditions) from ? ² _CDsem0 obtained as described above will be described. In this example, first, in order to ensure the accuracy of CD measurement reproducibility, parameter adjustment is performed to satisfy the condition of [Equation 11].

[Equation 11]

Required accuracy >σ ² _CDpattern0 ／(N ₁ ·N ₂ )+σ ² _CDsem0 ／((N ₁ ·N ₂ ·N ₃ ·Nframe)·(I ₀ ／Ip)

The measurement optimization unit 509 selects an appropriate combination of N ₃ and Nframe so that the condition of Equation 11 is satisfied. In addition, if the number of frames or the like is excessively increased, the throughput of the device decreases. Therefore, the smallest number of frames or the like that satisfy the condition of [Equation 11] may be selected. Further, by increasing the current Ip _,? ² _CDsem0 may be decreased. Conversely, if the beam current Ip is increased, the resolution of the device and the like decreases, so the lowest current that satisfies the condition of Equation 11 may be selected.

In addition, it is preferable to set the device conditions so that the measurement error is sufficiently small compared with the region 1 illustrated in FIG. 1, that is, the variation in pattern shape and dimension. In order to satisfy this condition, the measurement optimization unit 509 sets each parameter so that the condition of [Equation 12] is satisfied.

[Equation 12]

σ ² _{CDpattern 0} ／(N ₁ ·N ₂ )＞σ ² _CDsem0 ／((N ₁ ·N ₂ ·N ₃ ·Nframe)·(I ₀ ／I)

As with the fine hole pattern, in LCDU measurement when the shape-dimensional deviation of the measurement object is large, it is not necessary to set the measurement conditions in which the second term becomes small unnecessarily.

By performing the device adjustment based on the above calculation, it becomes possible to perform high-precision measurement in which measurement errors are suppressed.

In addition, when imaging the same FOV area with the number of frames (N), the two-dimensional signal intensity distribution for each scan of N scans is stored in the memory, and if they are appropriately added later, the number of frames (1 to N) is reduced. The imaging result can be acquired simultaneously. In this way, it is possible to significantly reduce the number of required imaging FOVs. That is, as illustrated in Fig. 10, if the number of frames is not changed in units of view, but information on the number of frames in one field of view can be obtained, the number of viewing positions required for imaging can be reduced, and throughput The improvement of becomes possible.

Next, the measurement reproducibility error is separated to estimate the true pattern dimension/shape deviation (LCDU). First, the same thing as the CD was performed for the LCDU, and conditions were set in area 1 of the LCDU. Under this setting condition, the obtained measurement result can be regarded as the standard deviation (σCD) of the observed CD deviation as the true value of the LCDU. However, in the case where the measurement reproducibility error cannot be ignored, the image analysis unit 508 calculates ? _CDpattern0 by [Equation 13] stored in the memory 506. σ _CDpattern0 becomes the estimated value of LCDU.

[Equation 13]

σ ² _{CDpattern 0} ＝N ₁ ·N ₂ ·σ ² _{CDobserved -σ} ² _CDsem0 ／(N ₃ ∙Nframe)·(I ₀ ／I)

In this way, the variance of the true pattern dimension shape deviation can be estimated from the variance (σ ² _observed ) of the CD and LCDU observation results when observed with an arbitrary number of frames (Nframe). Further, from the results, the dimensional deviation for a pattern of an arbitrary length can be estimated.

In addition, the setting of the device conditions can be set to a condition where the measurement error is sufficiently small compared to the deviation of the pattern shape and dimension, and in the LCDU measurement when the measurement object shape-dimensional deviation is large, such as a fine hole pattern, It is not necessary to set the measurement conditions in which the second term becomes small unnecessarily. As described above, in the present embodiment, it is possible to appropriately set imaging conditions or sampling plans that maximize throughput under conditions that satisfy the required measurement accuracy according to the measurement object.

The measurement analysis output unit 510 displays, for example, the number of frames, the beam current, and the like determined as described above on the display unit of the input device 504 or the workstation 622. The operator can see this indication and set the device conditions. Further, by displaying the number of frames or the like that satisfy the condition of [Equation 11] as candidates for setting conditions, the operator may be made to select the device condition. In addition, the measurement analysis output unit 510 displays the measurement result obtained by [Equation 13] on the display unit. At this time, not an index value such as a deviation, but a difference from a predetermined reference value may be displayed, or a chart for visually determining the degree of deviation such as a graph may be displayed.

Further, the number of frames obtained as described above may be set as the SEM apparatus condition.

[Example 2]

In the above-described embodiment, as the imaging unit, a scanning electron microscope, which is a kind of charged particle beam device, has been described as an example. For example, a focused ion beam device that generates an image by scanning a focused ion beam may be used as the imaging unit. The imaging unit independently stores the signal intensity distribution (image data) captured for each frame for a predetermined imaging area, and repeats this for a maximum number of frames as necessary to transfer each image data to the image storage unit. A plurality of image data are stored independently for an area.

An image processing unit included in an imaging unit such as a scanning electron microscope or a focused ion beam device, or included in a measurement device, combines the plurality of image data and synthesizes SEM images or SIM images of various frame numbers. The image analysis unit samples the measurement area from the image by the method described in Example 1, and fits the measurement and its deviation distribution by a predetermined model formula as a function of the number of samples or the number of frames.

The measurement optimization unit outputs an optimum imaging condition and a required sampling plan based on the above model equation by inputting boundary conditions such as required measurement reproduction accuracy and an upper limit on the number of frames. In addition, an estimated value of the expected measurement error is output. The sampling plan is reflected in measurement of subsequent wafers including the wafer.

In addition, when the SEM measurement reproducibility error exceeds a predetermined range, there is a possibility that the state of the SEM apparatus may deviate from the normal, and a predetermined warning is issued. In addition, when the pattern dimensional and shape deviation exceeds a predetermined range, there is a possibility that it may be an abnormality in the patterning process, and a predetermined warning is issued. The measurement analysis output unit removes the measurement error from the observed value and calculates a true dimension/shape deviation.

[Example 3]

In this embodiment, an example in which the LCDU or its distribution is correctly measured by the measurement method described in the first embodiment will be described. With respect to the same design pattern present in a plurality of locations on the wafer surface, the LCDU was measured by the method shown in Example 1, and fitted by the equation to obtain the true distribution of the dimensional and shape deviation within the wafer plane. By feeding back the obtained result to the process device, the above deviation was suppressed and the yield could be improved. In this example, only the true spatial deviation of the shape of the target pattern could be evaluated by subtracting the measurement reproducibility error.

501: SEM (scanning electron microscope) 502: pattern measuring device
503: design data storage medium 504: input device
505: arithmetic processing unit 506: memory
601: electron source 602: lead electrode
603: electron beam 604: condenser lens
605: scanning deflector 606: objective lens
607: vacuum sample chamber 608: sample stand
609: sample (wafer) 610: electron
611: secondary electron 612: conversion electrode
613: detector 614: control device
617: A/D converter 618: image processing unit
619: CPU 620: image memory
621: storage medium 622: workstation

Claims (6)

In a pattern measuring device including an arithmetic processing device that calculates a deviation of a measured value of a pattern to be measured based on a plurality of measured values acquired by a charged particle beam device,
A storage medium for storing the measured value of the pattern,
The arithmetic processing device is characterized in that σ _SEM0 representing a measurement reproducibility error is calculated using an equation including the following parameters and deviations (σ _observed ) of a plurality of measurement values formed at different positions stored in the storage medium. Pattern measurement device.
σ ² _observed ＝σ ² _{pattern 0} ／Np+σ ² _sem0 ／(Np·Nframe)
σ _pattern0 : Deviation due to the shape error of the pattern
Np: number of measurement points
Nframe: Value that changes according to device conditions The method of claim 1,
The Nframe is the number of frames of the charged particle beam device. The method of claim 2,
The arithmetic processing device obtains an Nframe satisfying a condition in which σ ² _pattern0 /Np+σ ² _sem0 /(Np·Nframe) is smaller than a required accuracy. The method of claim 1,
The pattern measurement device, wherein the arithmetic processing device _{obtains a} deviation (σ ² _{CDpattern 0} ) based on an equation having the following parameters.
σ ² _{CDpattern 0} ＝Np·σ ² _{CDobserved -σ} ² _CDsem0 ／Nframe
σ _CDpattern0 : deviation of the true pattern dimension
σ _CDobserved : Deviation of pattern dimension measurements
σ _CDsem0 : Deviation of pattern dimension measurement error The method of claim 1,
The plurality of measured values are values measured at different timings at different positions. In a storage medium storing a computer-readable computer program for causing a computer to obtain a deviation of a measured value of a pattern to be measured based on a plurality of measured values acquired by a charged particle beam device,
The program, the computer, characterized in that, by using an equation including the following parameters and deviations (σ _observed ) of a plurality of measurement values formed at different positions, to _obtain σ _SEM0 representing a measurement reproducibility error A storage medium storing a computer program readable by a computer.
σ ² _observed ＝σ ² _{pattern 0} ／Np+σ ² _sem0 ／(Np·Nframe)
σ _pattern0 : Deviation due to the shape error of the pattern
Np: number of measurement points
Nframe: Value that changes according to device conditions

Images